RESEARCH ARTICLE

Etiology and treatment of adrenoleukodystrophy: new insightsfrom DrosophilaHannah B. Gordon, Lourdes Valdez and Anthea Letsou*

ABSTRACTAdrenoleukodystrophy (ALD) is a fatal progressiveneurodegenerative disorder affecting brain white matter. The mostcommon form of ALD is X-linked (X-ALD) and results frommutation ofthe ABCD1-encoded very-long-chain fatty acid (VLCFA) transporter.X-ALD is clinically heterogeneous, with the cerebral form being themost severe. Diagnosed in boys usually between the ages of 4 and 8years, cerebral X-ALD symptoms progress rapidly (in as little as 2years) through declines in cognition, learning and behavior, toparalysis and ultimately to a vegetative state and death. Currently,there are no good treatments for X-ALD. Here, we exploit theDrosophila bubblegum (bgm) double bubble (dbb) model ofneurometabolic disease to expand diagnostic power andtherapeutic potential for ALD. We show that loss of the Drosophilalong-/very-long-chain acyl-CoA synthetase genes bgm and/or dbb isindistinguishable from loss of the Drosophila ABC transporter geneABCD. Shared loss-of-function phenotypes for synthetase andtransporter mutants point to a lipid metabolic pathway associationwith ALD-like neurodegenerative disease in Drosophila; a pathwayassociation that has yet to be established in humans. We also showthat manipulation of environment increases the severity ofneurodegeneration in bgm and dbb mutant flies, adding evenfurther to a suite of new candidate ALD disease-causing genes andpathways in humans. Finally, we show that it is a lack of lipidmetabolicpathway product and not (as commonly thought) an accumulation ofpathway precursor that is causative of neurometabolic disease:addition of medium-chain fatty acids to the diet of bgm or dbbmutantflies prevents the onset of neurodegeneration. Taken together, ourdata provide new foundations both for diagnosing ALD and fordesigning effective, mechanism-based treatment protocols.

This article has an associated First Person interview with the firstauthor of the paper.

KEY WORDS: Bubblegum (Bgm), Double bubble (Dbb), ABCD1,Fatty acid acyl-coA synthetase, Fatty acid transporter, Elongase,Neurodegeneration

INTRODUCTIONAdrenoleukodystrophy (ALD) is a neurodegenerative disorderassociated with mutation of the peroxisomal ABC transporter

protein ALDP (adrenoleukodystrophy protein) that is encoded bythe X chromosome locus ABCD1 (Mosser et al., 1993). Thespectrum of clinical features associated with ALDP mutation isbroad, ranging from adrenocortical insufficiency to slowlyprogressive myelopathy to cerebral demyelination (Raymondet al., 1999). ALDP is required for transport of very-long-chainfatty acids (VLCFAs), and ALDP deficits lead to VLCFAaccumulation in plasma and tissue (Engelen et al., 2012;Raymond et al., 1999). With respect to disease etiology, it isthought that VLCFA accumulation is toxic to the adrenal gland andto the myelin sheath that surrounds the many nerve cells of the body(Engelen et al., 2012). However, several inconsistencies exist inpatient studies that refute this model. First, while all individualsharboring disease-associated alleles of ABCD1 exhibit VLCFAlevel increases, some never manifest neurodegenerative symptoms(Engelen et al., 2012; Raymond et al., 1999). Second, VLCFAlevels do not correlate with patient neurological disabilities (Moser,1997). Third, although the two recipients of hematopoietic stem celltherapy showed improvement in their neurological symptoms,plasma VLCFA concentrations remained high (Cartier et al., 2009).In addition to the documented role for the ABC transporter inVLCFA metabolism and ALD, a role for the acyl-CoA synthases(ACSs) that function immediately upstream of ABC transporters infatty acid (FA) metabolism has long been contemplated, asdecreased ACS activity is another biochemical hallmark of ALD(Hashmi et al., 1986; Lazo et al., 1988; Wanders et al., 1988).Consistent with this idea is the recent identification of a patient withALD-like cerebral degeneration with a rare mutation in theSLC27a6-encoded ACS (Sivachenko et al., 2016).

The cerebral form of ALD is severely progressive and, in theabsence of cures and treatments, death is inevitable. It is also clearthat our current understanding of ALD disease etiology isinsufficient for the design of effective treatment protocols; in thisregard, therapeutic manipulation of VLCFA levels does not impactdisease progression (Engelen et al., 2012; Raymond et al., 1999).Animal models of neurometabolic disease, however, continue toenhance our understanding of ALD and yield new insights into ALDdiagnosis and management. In particular, neurometabolic diseasemodels in the mouse and fly are consistent with roles for ACSs inALD (Heinzer et al., 2003; Min and Benzer, 1999; Sivachenkoet al., 2016). These genetic platforms offer new opportunities for:(1) dissecting ALD-associated pathways enhancing early ALDdiagnostic power and (2) identifying molecular targets suitable fortherapeutic inhibition as well as alternative pathways that canpotentially be boosted to alleviate degeneration.

RESULTSLoss of neuronal Drosophila ABCD transporter functioncauses neurodegenerationAlthough neurodegeneration reminiscent of ALD has beensuccessfully modeled in ACS loss-of-function flies (Min andReceived 20 July 2017; Accepted 30 April 2018

Department of Human Genetics, University of Utah, Salt Lake City, UT 84112, USA.

*Author for correspondence (aletsou@genetics.utah.edu)

A.L., 0000-0002-8538-7886

This is an Open Access article distributed under the terms of the Creative Commons AttributionLicense (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use,distribution and reproduction in any medium provided that the original work is properly attributed.

1

© 2018. Published by The Company of Biologists Ltd | Disease Models & Mechanisms (2018) 11, dmm031286. doi:10.1242/dmm.031286

Disea

seModels&Mechan

isms

Benzer, 1999; Sivachenko et al., 2016), the analysis of orthologs ofthe X-linked ALD (X-ALD) human disease gene (ABCD1) inanimal models has remained elusive (Kobayashi et al., 1997) – invertebrates likely due to gene duplication. This said, the recentdevelopment of comprehensive and bioinformatically validatedRNAi libraries facilitating reverse genetic approaches has opened apipeline for gene validation in the non-redundant (less highlyduplicated) genome of the fly (Moulton and Letsou, 2016 for review).Using reciprocal BLASTp algorithms, we identified CG2316 as thesole Drosophila homolog of human ABCD1. CG2316, a fourthchromosome gene henceforth designated dABCD (for DrosophilaABCD), is 53% identical and 71% similar to human ABCD1(Fig. S1). Expression studies (Chintapalli et al., 2007) revealmoderate to high levels of dABCD in the adult head, consistentwith a role for dABCD in the maintenance of CNS health in flies.As no genetically defined lesions for dABCD exist, we employeda short-hairpin microRNA from the VALIUM20 collection (dsRNA-HMS02382; confirmed bioinformatically to have no off-targeteffects) to target dABCD for studies of gene function. We foundthat dABCDtub>dsRNA transgenics survive to adulthood, but sufferfrom neurodegeneration. Specifically, dABCDtub>dsRNA flies exhibit abrain phenotype indistinguishable from that of animals homozygousfor amorphic alleles of the bgm- and dbb-encoded Drosophilalong/very-long-chain ACSs – that being an age-dependent retinaldisorganization that is distinguished by retinal holes and pigmentcell loss (Fig. 1A,B,F). Previous reports (Min and Benzer, 1999;Sivachenko et al., 2016) have validated both bgm and dbb asALD-like models of neurodegeneration. However, the utility of theselines as disease models is now further bolstered by their sharedloss-of-function phenotype with dABCD. Moreover, it is clearthat ABCD1 should be added to the growing list of humanneurodegenerative-disease-related genes with functional homologsin Drosophila (Chien et al., 2002).Having established here and elsewhere (Sivachenko et al., 2016)

that both neurons and glia die in Drosophila ABCD and ACS

models of neurodegeneration, we next sought to determine the cell-type-specific requirements for the Drosophila-encoded FAtransporter and synthetases. To this end, we used elav (neuronal),sim (embryonic midline glial and adult neuronal) and repo (glial)drivers (FlyBase, 2003) to mediate cell-type-specific expression ofdsRNA targeting dABCD. Neuronal disruption of dABCD indABCDelav>dsRNA and dABCDsim>dsRNA transgenics recapitulatesretinal defects seen in dABCDtub>dsRNA animals; in contrast, nodefects result from glial-specific disruption in dABCDrepo>dsRNA

transgenics (Fig. 1C-E,K). Complementing this analysis is our studyof the effects of cell-type-specific expression of a Drosophila bgm+

transgene in a bgm1 null mutant background. Both ubiquitous andneuronal bgm+ expression rescue age-dependent neurodegenerationin bgm1 mutants, although glial expression does not (Fig. 1F-J,L).Our observation of identical tissue-specific effects for dsRNA-HMS02382-mediated dABCD gene disruption and bgm+ generescue is consistent with bioinformatic exclusion of off-targeteffects for dsRNA-HMS02382. Moreover, results from targeteddisruption (dABCD) and rescue (bgm) studies show that the primarysite of ABCD/ACS function in the Drosophila model of ALD-likeneurodegeneration is the neuron and point to this cell type as theoptimal target for prevention of neurodegeneration in ACS andABCD fly models, and, by extension, for ALD therapy in humans.

Gene-environment interactions modulate ALD penetranceand expressivityAs is true for ALD patients, neurodegeneration in bgm, dbb anddABCD flies is incompletely penetrant and variably expressed(Sivachenko et al., 2016; see also Fig. 1). The origin of thisvariability is so far unknown, although multiple reports havepointed to an environmental interaction (Weller et al., 1992;Raymond, et al., 2010; Sivachenko et al., 2016). Thus, to examineenvironmental contributions to phenotype, we examined responsesof bgm1 and dbb1 amorphs to environmental stress in the form oflight manipulation (Johnson et al., 2002). In contrast to animals

Fig. 1. dABCD and bgm are required in adult retinalneurons. (A) Retinal cross-sections of UAS-dABCD-RNAicontrol animals (no Gal4 driver) reveal a highly organizedommatidial structure. (B) In contrast, ubiquitous knockdownof dABCD results in a shared loss-of-function phenotype withbgm and dbbmutants, with holes and disrupted pigment cellsbetween ommatidia. (C,E) Neuronal knockdown of dABCD,but not glial knockdown (D), also leads to neurodegeneration.(F) Representative cross-section of a bgm mutant at day 20post-eclosion. (G) Ubiquitous expression (tub-Gal4) of aninducible bgm+ transgene rescues neurodegeneration in abgm mutant background. (H) Driving bgm+ expression inneuronal cells (sim-Gal4) is sufficient to rescue the mutantphenotype; however, (I) driving bgm+ in glial cells (repo-Gal4)does not result in rescue. (J)DJ667-Gal4, which was used asa negative control as it is specifically expressed in flightmuscles, fails to rescue. (K) Blinded quantification ofneurodegenerative (ND) phenotypes in A-E. (L) Blindedquantification of neurodegenerative phenotype seen in G-J.In K and L, scores were compared by one-way ANOVA withDunnett’s test and P-values. Data points represent biologicalreplicates and are presented with means±s.e.m. ***P<0.001.Arrowheads point to areas of retinal degeneration.

2

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm031286. doi:10.1242/dmm.031286

Disea

seModels&Mechan

isms

raised in normal light/dark conditions (12 h light:12 h dark),animals raised in constant dark (24 h dark) exhibit significantlyless neurodegeneration; in the case of dbb1, neurodegenerationappears to be blocked entirely (Fig. 2A-F,J). In complementaryconstant light conditions, we observed a significant exacerbationof neurodegeneration in all backgrounds, including the wildtype (Fig. 2G-J). Finally, in defining constant-light-inducedneurodegeneration in wild-type animals as baseline, wedetermined that enhanced neurodegeneration in bgm1 and dbb1

animals is synergistic (Fig. 2J). Thus, environmental stress in theform of light modifies neurodegenerative phenotypes in bgmand dbb models of neurodegeneration, demonstrating that gene-environment interactions can modulate penetrance and expressivityof neurodegenerative phenotypes.

Product insufficiency is causative of ALDDespite accumulation of VLCFAs (ACS and ABCD1 substrates)in ALD patients, it is still debated whether this marker of disease is

also causative of disease, and whether accumulating VLCFAsshould serve as a therapeutic target (Raymond et al., 1999). Here,we consider the alternative possibility – that it is the absence ofactivated VLCFAs and/or their metabolic products that iscausative of disease. As a first step toward disease therapeuticsand in our first direct test of the lack of product hypothesis forALD, we fed bgm1 and dbb1 animals a diet high in medium-chainfatty acids [7% coconut oil (Birse et al., 2010)] from day 0 (d0) today 20 (d20) post-eclosion, anticipating that bypass of the geneticblock to activating long-chain FAs (LCFAs)/VLCFAs in bgmand dbb mutants via the elongase pathway might suppressneurodegeneration (Fig. 3A). Indeed, supplementation of bgm1

and dbb1 mutant diets with medium-chain FAs significantlyreduces retinal defects observed at d20 post-eclosion in both bgm1

and dbb1 animals (Fig. 3B-D,H-J,Q). Second, in anticipation thatincreasing LCFAs/VLCFAs will enhance neurodegeneration ifaccumulating precursors are toxic (see Fig. 3A), we examinedneurodegeneration in bgm1 and dbb1 animals fed a diet high inLCFAs (described in Carvalho et al., 2012). We found no changesin neurodegeneration in animals fed LCFA-enriched diets(Fig. 3E-G,P). That mutant neurodegenerative phenotypes wererescued by medium-chain-FA dietary supplementation and notexacerbated by LCFA dietary supplementation points to productloss as being causative of neurodegenerative disease.

A parallel route to VLCFA production (the elongase pathway)is required for CNS health and maintenanceGiven the deficiencies of Lorenzo’s oil in the treatment of ALDpatients (Moser, 1999; van Geel et al., 1999; Zinkham et al., 1993)and as an extension of our medium-chain dietary treatment results inbgm and dbb flies, we next tested whether disruption of the FAelongase pathway that uses activated medium-chain FAs to produceactivated LCFAs is associated with neurodegeneration (see Fig. 3A).Using BLASTp, we identified four genes encoding Drosophilaelongases [there are seven in humans (Jump, 2009)], but only one,CG2781 (44-56% identical to ELOVL1, 7 and 4 and henceforthidentified as dELOVL forDrosophila ELOVL; Fig. S2), is expressed ina spatial and temporal manner analogous to dABCD and predicative ofa role in neuronal health and maintenance (Chintapalli et al., 2007).There are no genetically defined dELOVL mutants; thus, as we didpreviously for dABCD1, we used the binary UAS-GAL4 system incombination with RNAi-mediated gene disruption methods. We usedtwo reagents to disrupt ELOVL gene function and thereby assessits function in CNS health and maintenance. The first, dsRNA-HMC03112, is from the VALIUM20 TRiP collection and has beenconfirmed bioinformatically to have no off-target effects. The second,dsRNA-GD16713, is from the Vienna Drosophila Resource Center.While ubiquitous dELOVL disruption (in dELOVLtub>dsRNA-HMC03112

transgenics) leads to lethality before eclosion (data not shown)and points to an early essential role for dELOVL, specific neuronalknockdown of dELOVL (in dELOVLsim>dsRNA-HMC03112 anddELOVLelav>dsRNA-HMC03112) leads to neurodegeneration. Animalsexhibit retinal defects, including holes and lost pigment cells,phenotypes replicating those that we have observed in bgm-, dbb-and dABCD1-deficient animals. In an extension of the analysis,targeted disruption of dELOVL in glia (in dELOVLrepo>dsRNA-HMC03112

transgenics) does not produce degenerative phenotypes (Fig. 3K-O).Thus, cell-type specificity for all pathway components in VLCFAmetabolism (ACSs, elongases and transporters) is neuronal. Despitea high background for leaky dsRNA-GD16713 expression thatcauses death in the absence of a GAL4 driver, we reproduced alldsRNA-HMC03112 qualitative results using dsRNA-GD16713

Fig. 2. Light-dark cycles modulate neurodegeneration in bgm and dbbsingle mutants. (A-C) Cross-sections of retinas from wild-type (wt), bgm anddbb animals raised in standard 12-h light/dark cycles (L/D), with bgm and dbbmutants showing normal patterns of degeneration. (D-F) Constant dark (‘D’)alleviates degeneration, whereas (G-I) constant light (‘L’) exacerbatesdegeneration. (J) Blinded quantification of neurodegenerative (ND)phenotypes. Scores were compared using one-way ANOVA with Dunnett’s.Data points represent biological replicates and are presented withmeans±s.e.m. ***P<0.001. Arrowheads point to areas of retinal degeneration.

3

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm031286. doi:10.1242/dmm.031286

Disea

seModels&Mechan

isms

(data not shown). Together, these data show that product loss iscausative of disease and that mutations in genes encoding elongasesand medium-chain FA acyl-CoA synthetases should be consideredcandidate ALD-causing disease genes, as well as targets fortherapeutic options.

DISCUSSIONALD is a progressive neurodegenerative disease, with the mostsevere form claiming the lives of school-age boys. AlthoughABCD1, the gene responsible for the most common form of ALD(X-ALD), has been cloned, the etiology of the disease has remainedelusive and there are still no satisfactory treatments or cures (forreview see Gordon et al., 2014). Using in vivo models ofneurometabolic disease, we here provide new insights into humanALD. Our demonstration that mutations in long- and medium-chainFA metabolic pathways in Drosophila yield shared loss-of-functionneurodegenerative phenotypes extends a single-gene association(ABCD1) for ALD to a pathway association (lipid metabolism).

Importantly, this more expansive view of ALD offers a possibility ofdiagnosis to some of the 50% of leukodystrophy patients withundiagnosed conditions (Gordon et al., 2014). In addition, ourdemonstration that neurodegeneration in fly models of ALD doesnot result from a buildup of FA precursors, but instead is caused by alack of activated FA product, shifts our understanding of ALDetiology and is expected to have profound effects on the design ofeffective therapeutics. Indeed, our data indicate that a diet high inmedium-chain FAs provides a potential therapeutic approach forleukodystrophy patients with ACS mutations (Sivachenko et al.,2016). At the very least, our study validates the continued search forremedies other than Lorenzo’s oil for the treatment of X-ALD(Eichler et al., 2017; Kolata, 2017).

Although hundreds of ABCD1 alleles are associated with ALD,no genotype-phenotype correlations have emerged. Likewise,inheritance of the same allele within kindreds (or even an allelecommon to monozygotic twins) can lead to different diseasephenotypes (Berger et al., 1994; Korenke et al., 1996). Thus, it

Fig. 3. Neurodegeneration in bgm and dbb animals iscaused by a lack of activated fatty acid (FA)product(s). (A) Two pathways, one depending on long-and the other on medium-chain FAs, converge toproduce activated long- and very-long-chain fatty acidproducts. (B-D) Cross-sections of retinas from wild-type(wt), bgm and dbb animals fed a standard diet (SD).(E-G) Retinas from wt, bgm and dbb animals fed a dietenriched in long-chain FAs (LCD) are indistinguishablefrom genotypically matched animals fed SD. (H-J) Dietsrich in medium-chain fatty acids (MCD) suppressneurodegeneration in bgm and dbb mutants.(P,Q) Blinded scoring of neurodegeneration in animalsfed SD compared with those fed with LCD or MCD.(K-N) Cross-sections of retinas from animals with andwithout tissue-specific expression of an RNAi targetingtheDrosophila elongaseCG2781. (O) Blinded scoring ofneurodegenerative (ND) phenotypes in the RNAiexperiment. All scores were compared by one-wayANOVA with Dunnett’s test. Data points representbiological replicates and are presented with means±s.e.m. **P<0.01, *P<0.05. Arrowheads point to areas ofretinal degeneration.

4

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm031286. doi:10.1242/dmm.031286

Disea

seModels&Mechan

isms

seems clear that gene-environment interactions modulate ALDpenetrance and expressivity. Gene-environment interactions extendto other players in the lipid metabolic pathway associated with ALDas well. In this regard, each of the young brothers in a recentlydescribed Utah leukodystrophy family harbors an allele pairassociated with two incompletely penetrant conditions (PRRT2/childhood epilepsy and Slc27a6/leukodystrophy). Only the youngerbrother, however, exhibits both seizure and leukodystrophyphenotypes (Sivachenko et al., 2016).Our finding that environmental stress in the form of light

modifies neurodegenerative phenotypes in bgm and dbb models ofneurodegeneration provides direct evidence for a gene-environmentinteraction that modulates penetrance and expressivity ofneurodegenerative phenotypes (see Fig. 2). These data bolster ourview that: (1) environmental stress in the form of seizure triggeredneurodegeneration in a leukodystrophy patient with a predisposingACS mutation (Sivachenko et al., 2016), and (2) traumatic braininjury triggered neurodegeneration in patients with catastrophicpresentations of ALD (Raymond et al., 2010; Vawter-Lee et al.,2015; Weller et al., 1992). Moreover, as stress might precipitate adegenerative cellular phenotype when product generation isrequired to repair damaged or depleted cellular components, ourbgm/dbb stress studies suggest that esterified LCFA and VLCFAproducts of Bgm and Dbb activity are required to preventneurodegeneration, a hypothesis that contradicts the long-heldview that precursor accumulation is causative of neurodegenerationin ALD patients.In the Drosophila bgm/dbb model of neurodegeneration,

inclusions in brain tissue and elevated levels of VLCFAsprovide clear diagnostic markers of disease, and point to flymutants as powerful in vivo models for testing the effects of FAdysregulation on neurodegeneration (Sivachenko et al., 2016). Inhumans, evidence for accumulating FAs as being causative ofdisease comes primarily from tissue culture studies, whereaccumulating FAs can lead to cell death (Hein et al., 2008;Reiser et al., 2006; Ulloth et al., 2003). Contradicting this view,however, are data from ABCD1 hemizygotes showing that,although all individuals harboring mutant alleles of ABCD1exhibit similar significant increases in their circulating VLCFAlevels, these correlate with neither the development of disease northe timing of disease onset (Dubey et al., 2005; Raymond et al.,1999). Moreover, some recipients of hematopoietic stem cell genetherapy show improvement in neurological symptoms despiteplasma VLCFA concentrations remaining significantly high(Cartier et al., 2009). Understanding disease etiology representsan essential first step in providing appropriate therapies. Thecurrent therapeutic option for the most severe ALD cases is bonemarrow transplant; however, the procedure itself carriessubstantial risk and is not always successful (Berger et al., 2010;Gordon et al., 2014). Another touted therapy called Lorenzo’soil is thought to target accumulated VLCFAs but remains acontroversial option, with most agreeing that it is ineffective(Aubourg et al., 1993; Berger et al., 2010; Gordon et al., 2014).Here, we demonstrate that bgm and dbb neurodegenerative

phenotypes are rescued by medium-chain FA dietarysupplementation, while being unaffected by LCFA (see Fig. 3).Together, data from these complementary studies identify productloss as causative of neurodegenerative disease. Indeed, end productsof peroxisomal metabolism (including both glycerolipids andplasmalogens) constitute more than 80% of the phospholipidcontent of brain white matter (Schrader and Fahimi, 2008).Additionally, our results are the first to highlight potential for the

activated medium-chain elongation pathway as an alternate routeto production of missing VLCFA product(s) in ACS mutants.Of course, dietary supplementation is not necessarily expectedto rescue ABCD Drosophila mutants (or to provide therapeuticvalue to X-ALD patients) because dABCD/ABCD functions at theintersection of the long/very-long- and medium-chain lipidmetabolic pathways (see Fig. 3A). This said, our prior associationof the SLC27a6-encoded ACS with ALD (Sivachenko et al., 2016)suggests that there are forms of the disease that are likely to bealleviated by dietary supplementation with medium-chain FAs.

Finally, over half of leukodystrophies remain undiagnosed. Ouridentification of the elongase pathway identifies new candidates fordisease genes. At the biochemical level, fatty acyl-CoA chainelongation involves the addition of two carbon units to an existingfatty acyl-CoA, thereby bypassing a requirement for VLCFAactivation by long- and very-long-chain ACSs (Beaudoin et al.,2002). Elongases have been implicated in ALD etiology, as enzymelevels are increased in induced pluripotent stem cell (IPSC)-derivedbrain cells from ALD patients (Baarine et al., 2015). Althoughincreased elongase levels have been interpreted to mean thatelongase function might exacerbate disease in patients, functionaltests have yet to be undertaken and another interpretation of the datais that elongase levels are upregulated in ABCD1 mutants as analternate route to the production of essential activated VLCFAs.Interestingly, in humans, the longest VLCFA species are found onlyin tissues expressing elongases, namely the retina, brain and testis,all three of which are key tissues affected in ALD (Agbaga et al.,2010; Ahmad et al., 2009).

In summary, ALD in its most severe form results in acute andrapidly progressing degeneration of brain white matter and leads todeath within a few years of diagnosis. The etiology of ALD hasremained poorly understood, and there is still no treatment for thiscondition. ALD has long been thought to result from a buildupof VLCFAs in the brains of affected individuals. We used theDrosophila model of neurodegeneration to show that this long-heldview is incorrect. While accumulating VLCFAs are indeed a markerof neurodegenerative disease in both flies and humans, we show that itis the absence of activated VLCFAs and or/their metabolic productsthat is causative of disease in the fly model. Our studies contributeto the fields of ALD and neurodegenerative disease in three majorways: (1) we show that activated VLCFAs and/or their products arenecessary for neuronal health and maintenance; (2) we identify newcandidate ALD-disease-causing genes, and (3) we show that a diethigh in medium-chain FAs shows promise as a potential therapeuticapproach for patients with neurometabolic degenerative disease.

MATERIALS AND METHODSDrosophila stocksGal4 lines used to drive targeted transgene expression include repo-Gal4(BL#7415), elav-Gal4 (BL#8760), sim-Gal4 (BL#9150) and DJ667-Gal4(BL#8171); tubulin-Gal4 (w1118; P{tubP-gal4}LL7/TM3, P{Dfd-GMR-nvYFP}3 Sb1) was the gift of MarkMetzstein (University of Utah, Salt LakeCity, UT). The bgm1 and dbb1 mutants have been described (Min andBenzer, 1999; Sivachenko et al., 2016). dsRNA lines targeting dABCD1 anddELOVL were obtained from the Transgenic RNAi Project at HarvardMedical School (#41984 and #50710, respectively); an additional dsRNAline targeting dELOVL was obtained from the Vienna Drosophila ResourceCenter (v102543). Unless otherwise noted, all flies were raised on a standardcornmeal diet at 25°C with 12-h light:dark cycles.

For tissue-specific expression of bgm, its coding sequence was insertedinto pFLAG-CMV-5a (Sigma #E7523) using primers 5′-ATAAAGCTT-ATGTCCACGATAGACGCGCTC-3′ and 5′-GCGGTACCGGCATATA-GTTTCTCGATCTC-3′. The tagged version of the gene was subsequently

5

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm031286. doi:10.1242/dmm.031286

Disea

seModels&Mechan

isms

inserted into the Drosophila expression vector pUAST using primers 5′-AATGGGCGGTAGGCGTGTACG-3′ and 5′-AATCTAGACTCGAGAT-TAGGACAAGGCTGGTGGGC-3′. pUAS-bgm-FLAGwas sequence veri-fied and co-injected with transposase Δ2-3 into dechorionated nosΦC31; +;vk27 embryos 1-2 h after egg lay. G0 injected animals were mated to w1118,and progeny were screened for transformed germlines based on eye color.The UAS-bgm-FLAG line used for our studies is homozygous viable, withthe transgene insertion on the second chromosome.

Diet and light manipulationsMedium- and long-chain diets were prepared as previously described (Birseet al., 2010; Carvalho et al., 2012). Adult males were collected within 24 hof eclosion and maintained on prescribed diets until sacrifice 20-22 dayspost-eclosion. For light manipulations, males were isolated within 48 h ofeclosion and habituated in 24 h light, 12 h light/dark, or 24 h dark cycles in atemperature- and humidity-controlled room until sacrifice at 20-22 dayspost-eclosion.

Aging and histologyHeads from adultDrosophilamales were prepared, sectioned and imaged aspreviously described (Sivachenko et al., 2016). Samples were scored blindlyin three to five serial sections for each animal. The degree of retinaldegeneration was scored qualitatively as 0 for normal appearance, 1 for mildtissue loss, 2 for moderate degeneration and 3 for severe degeneration (Caoet al., 2013). Data were analyzed by ANOVA andWelch two-sample t-tests.Data were analyzed using GraphPad Prism software (GraphPad Software).

AcknowledgementsThe authors thank Suzanne Kimball for technical support, Kyung-Tai Min and MarkMetzstein for fly lines, Diana Lim for figure preparation, and Gab Kardon and MarkMetzstein for comments on the manuscript.

Competing interestsThe authors declare no competing or financial interests.

Author contributionsConceptualization: H.B.G., A.L.; Methodology: H.B.G., A.L.; Validation: H.B.G., A.L.;Formal analysis: H.B.G., A.L.; Investigation: H.B.G., L.V.; Resources: H.B.G., A.L.;Data curation: H.B.G.; Writing - original draft: H.B.G., A.L.; Writing - review & editing:H.B.G., A.L.; Supervision: A.L.; Project administration: A.L.; Funding acquisition: A.L.

FundingThis research was supported by a grant from theNational Institutes of Health (NIH) toA.L. (R01NS065474). The authors also acknowledge training support from theHoward Hughes Medical Institute (HHMI) and Little Red Riding Hood ResearchFoundations (H.G.).

Supplementary informationSupplementary information available online athttp://dmm.biologists.org/lookup/doi/10.1242/dmm.031286.supplemental

ReferencesAgbaga, M.-P., Mandal, M. N. A. and Anderson, R. E. (2010). Retinal very long-chain PUFAs: new insights from studies on ELOVL4 protein. J. Lipid Res. 51,1624-1642.

Ahmad, A., Mazhar, A. U. and Anwar, M. (2009). Farber disease: a rareneurodegenerative disorder. J. Coll. Physicians Surg. Pak. 19, 67-68.

Aubourg, P., Adamsbaum, C., Lavallard-Rousseau, M.-C., Rocchiccioli, F.,Cartier, N., Jambaque, I., Jakobezak, C., Lemaitre, A., Boureau, F., Wolf, C.et al. (1993). A two-year trial of oleic and erucic acids (“Lorenzo’s oil”) as treatmentfor adrenomyeloneuropathy. N. Engl. J. Med. 329, 745-752.

Baarine, M., Khan, M., Singh, A. and Singh, I. (2015). Functional characterizationof IPSC-derived brain cells as a model for X-linked adrenoleukodystrophy. PLoSONE 10, e0143238.

Beaudoin, F., Gable, K., Sayanova, O., Dunn, T. and Napier, J. A. (2002). ASaccharomyces cerevisiae gene required for heterologous fatty acid elongaseactivity encodes a microsomal beta-keto-reductase. J. Biol. Chem. 277,11481-11488.

Berger, J., Molzer, B., Fae, I. and Bernheimer, H. (1994). X-linkedadrenoleukodystrophy (ALD): a novel mutation of the ALD gene in 6 membersof a family presenting with 5 different phenotypes. Biochem. Biophys. Res.Commun. 205, 1638-1643.

Berger, J., Pujol, A., Aubourg, P. and Forss-Petter, S. (2010). Current and futurepharmacological treatment strategies in X-linked adrenoleukodystrophy. BrainPathol. 20, 845-856.

Birse, R. T., Choi, J., Reardon, K., Rodriguez, J., Graham, S., Diop, S., Ocorr, K.,Bodmer, R. and Oldham, S. (2010). High-fat-diet-induced obesity and heartdysfunction are regulated by the TOR pathway in Drosophila. Cell Metab. 12,533-544.

Cao, Y., Chtarbanova, S., Petersen, A. J. and Ganetzky, B. (2013). Dnr1mutations cause neurodegeneration in Drosophila by activating the innateimmune response in the brain. Proc. Natl. Acad. Sci. USA 110, E1752-E1760.

Cartier, N., Hacein-Bey-Abina, S., Bartholomae, C. C., Veres, G., Schmidt, M.,Kutschera, I., Vidaud, M., Abel, U., Dal-Cortivo, L., Caccavelli, L. et al. (2009).Hematopoietic stem cell gene therapy with a lentiviral vector in X-linkedadrenoleukodystrophy. Science 326, 818-823.

Carvalho, M., Sampaio, J. L., Palm, W., Brankatschk, M., Eaton, S. andShevchenko, A. (2012). Effects of diet and development on the Drosophilalipidome. Mol. Syst. Biol. 8, 600.

Chien, S., Reiter, L. T., Bier, E. and Gribskov, M. (2002). Homophila: humandisease gene cognates in Drosophila. Nucleic Acids Res. 30, 149-151.

Chintapalli, V. R., Wang, J. and Dow, J. A. T. (2007). Using FlyAtlas to identifybetter Drosophila melanogaster models of human disease. Nat. Genet. 39,715-720.

Dubey, P., Raymond, G. V., Moser, A. B., Kharkar, S., Bezman, L. and Moser,H. W. (2005). Adrenal insufficiency in asymptomatic adrenoleukodystrophypatients identified by very long-chain fatty acid screening. J. Pediatr. 146,528-532.

Eichler, F., Duncan, C., Musolino, P. L., Orchard, P. J., De Oliveira, S., Thrasher,A. J., Armant, M., Dansereau, C., Lund, T. C. and Miller, W. P. (2017).Hematopoietic stem-cell gene therapy for cerebral adrenoleukodystrophy.N. Engl. J. Med. 377, 1630-1638.

Engelen, M., Kemp, S., de Visser, M., van Geel, B. M., Wanders, R. J. A.,Aubourg, P. and Poll-The, B. T. (2012). X-linked adrenoleukodystrophy (X-ALD):clinical presentation and guidelines for diagnosis, follow-up and management.Orphanet J. Rare Dis. 7, 51.

FlyBase (2003). The Flybase database of the Drosophila genome projects andcommunity literature. Nucleic Acid Res. 31, 172-175.

Gordon, H. B., Letsou, A. and Bonkowsky, J. L. (2014). The leukodystrophies.Semin. Neurol. 34, 312-320.

Hashmi, M., Stanley, W. and Singh, I. (1986). Lignoceroyl-CoASH ligase: enzymedefect in fatty acid beta-oxidation system in X-linked childhoodadrenoleukodystrophy. FEBS Lett. 196, 247-250.

Hein, S., Schonfeld, P., Kahlert, S. and Reiser, G. (2008). Toxic effects of X-linkedadrenoleukodystrophy-associated, very long chain fatty acids on glial cells andneurons from rat hippocampus in culture. Hum. Mol. Genet. 17, 1750-1761.

Heinzer, A. K., Watkins, P. A., Lu, J. F., Kemp, S., Moser, A. B., Li, Y. Y., Mihalik,S., Powers, J. M. and Smith, K. D. (2003). A very long-chain acyl-CoAsynthetase-deficient mouse and its relevance to X-linked adrenoleukodystrophy.Hum. Mol. Genet. 12, 1145-1154.

Johnson, K., Grawe, F., Grzeschik, N. and Knust, E. (2002). Drosophila crumbs isrequired to inhibit light-induced photoreceptor degeneration. Curr. Biol. 12,1675-1680.

Jump, D. B. (2009). Mammalian fatty acid elongases. Methods Mol. Biol. 579,375-389.

Kobayashi, T., Shinnoh, N., Kondo, A. and Yamada, T. (1997).Adrenoleukodystrophy protein-deficient mice represent abnormality of very longchain fatty acid metabolism. Biochem. Biophys. Res. Commun. 232, 631-636.

Kolata, G. (2017). In a First, Gene Therapy Halts a Fatal Brain Disease. The NewYork Times Oct. 5, 2017.

Korenke, G. C., Fuchs, S., Krasemann, E., Doerr, H. G., Wilichowski, E.,Hunneman, D. H. and Hanefeld, F. (1996). Cerebral adrenoleukodystrophy(ALD) in only one of monozygotic twins with an identical ALD genotype. Ann.Neurol. 40, 254-257.

Lazo, O., Contreras, M., Hashmi, M., Stanley, W., Irazu, C. and Singh, I. (1988).Peroxisomal lignoceroyl-CoA ligase deficiency in childhoodadrenoleukodystrophy and adrenomyeloneuropathy. Proc. Natl. Acad. Sci. USA85, 7647-7651.

Min, K. T. and Benzer, S. (1999). Preventing neurodegeneration in the Drosophilamutant bubblegum [see comments]. Science 284, 1985-1988.

Moser, H. W. (1997). Adrenoleukodystrophy: phenotype, genetics, pathogenesisand therapy. Brain 120, 1485-1508.

Moser, H. W. (1999). Treatment of X-linked adrenoleukodystrophy with Lorenzo’soil. J. Neurol. Neurosurg. Psychiatry 67, 279-280.

Mosser, J., Douar, A.-M., Sarde, C.-O., Kioschis, P., Feil, R., Moser, H., Poustka,A.-M., Mandel, J.-L. and Aubourg, P. (1993). Putative X-linkedadrenoleukodystrophy gene shares unexpected homology with ABCtransporters. Nature 361, 726-730.

Moulton, M. J. and Letsou, A. (2016). Modeling congenital disease andinborn errors of development in Drosophila melanogaster. Dis. Model. Mech. 9,253-269.

6

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm031286. doi:10.1242/dmm.031286

Disea

seModels&Mechan

isms

Raymond, G. V., Moser, A. B. and Fatemi, A. (1999). X-LinkedAdrenoleukodystrophy. In GeneReviews® [Internet] (ed. M. P. Adam, H. H.Ardinger, R. A. Pagon, et al.). Seattle: University of Washington (1993-2018).

Raymond, G. V., Seidman, R., Monteith, T. S., Kolodny, E., Sathe, S., Mahmood,A. and Powers, J. M. (2010). Head trauma can initiate the onset of adreno-leukodystrophy. J. Neurol. Sci. 290, 70-74.

Reiser, G., Schonfeld, P. and Kahlert, S. (2006). Mechanism of toxicity of thebranched-chain fatty acid phytanic acid, amarker of Refsum disease, in astrocytesinvolves mitochondrial impairment. Int. J. Dev. Neurosci. 24, 113-122.

Schrader, M. and Fahimi, H. D. (2008). The peroxisome: still a mysteriousorganelle. Histochem. Cell Biol. 129, 421-440.

Sivachenko, A., Gordon, H. B., Kimball, S. S., Gavin, E. J., Bonkowsky, J. L. andLetsou, A. (2016). Neurodegeneration in a Drosophila model ofadrenoleukodystrophy: the roles of the Bubblegum and Double bubble acyl-CoAsynthetases. Dis. Model. Mech. 9, 377-387.

Ulloth, J. E., Casiano, C. A. andDe Leon,M. (2003). Palmitic and stearic fatty acidsinduce caspase-dependent and -independent cell death in nerve growth factordifferentiated PC12 cells. J. Neurochem. 84, 655-668.

van Geel, B. M., Assies, J., Haverkort, E. B., Koelman, J. H., Verbeeten, B., Jr,Wanders, R. J. A. and Barth, P. G. (1999). Progression of abnormalities inadrenomyeloneuropathy and neurologically asymptomatic X-linkedadrenoleukodystrophy despite treatment with “Lorenzo’s oil”. J. Neurol.Neurosurg. Psychiatry 67, 290-299.

Vawter-Lee, M. M., Hallinan, B. E., Burrow, T. A., Spaeth, C. G. and Arthur, T. M.(2015). A novel catastrophic presentation of X-linked Adrenoleukodystrophy.JIMD Rep. 24, 97-102.

Wanders, R. J. A., van Roermund, C.W. T., Schutgens, R. B. H., van den Bosch,H. and Tager, J. M. (1988). Impaired ability of peroxisomes to activate very-long-chain fatty acids in X-linked adrenoleukodystrophy. Lancet 332, 170.

Weller, M., Liedtke, W., Petersen, D., Opitz, H. and Poremba, M. (1992). Very-late-onset adrenoleukodystrophy: possible precipitation of demyelination bycerebral contusion. Neurology 42, 367-370.

Zinkham, W. H., Kickler, T., Borel, J. and Moser, H. W. (1993). Lorenzo’s oil andthrombocytopenia in patients with adrenoleukodystrophy. N. Engl. J. Med. 328,1126-1127.

7

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm031286. doi:10.1242/dmm.031286

Disea

seModels&Mechan

isms